{ "name": "AI servers by OEM and ODM", "url": "https://aiserverlist.com/", "license": "CC BY 4.0", "cite": "AI Server List, https://aiserverlist.com/", "updated": "2026-11-11", "fields": [ { "key": "maker", "label": "Maker", "type": "string", "description": "The OEM, ODM or system builder that sells the server model. Corporate names follow the vendor’s own pages.", "filterable": false, "sortable": true }, { "key": "gpu_vendor", "label": "GPU vendor", "type": "enum", "description": "Maker of the accelerators named in the GPU model field: NVIDIA, AMD or Intel. Blank when the vendor page names no accelerator.", "filterable": true, "sortable": true, "enumValues": [ "NVIDIA", "AMD", "Intel" ] }, { "key": "gpu_model", "label": "GPU or accelerator", "type": "string", "description": "Accelerators the vendor page names for this model. A model that supports several options lists the option that was read, marked as one supported option.", "filterable": false, "sortable": true }, { "key": "gpu_count", "label": "GPUs per server", "type": "number", "unit": "GPUs", "description": "Installed count or supported maximum per server, as the qualifier in the source says. Rack-scale systems show the GPUs in the whole rack (72 for an NVL72). Blank when the page gives no count.", "filterable": true, "sortable": true }, { "key": "rack_units", "label": "Chassis height", "type": "number", "unit": "U", "description": "Rack height of the server as the vendor states it. Open Rack Units (OU) are not the same as EIA U and sit in the form factor field instead.", "filterable": true, "sortable": true }, { "key": "form_factor", "label": "Form factor", "type": "string", "description": "Configuration-specific chassis or rack description where the vendor splits it by cooling or rack standard.", "filterable": false, "sortable": true }, { "key": "cooling_type", "label": "Cooling type", "type": "enum", "description": "Air, liquid, or air or liquid by configuration, read from the vendor’s cooling statement. Blank when the page does not state cooling.", "filterable": true, "sortable": true, "enumValues": [ "Air", "Liquid", "Air or liquid" ] }, { "key": "cooling", "label": "Cooling (vendor wording)", "type": "string", "description": "The vendor’s cooling statement with configuration qualifiers kept.", "filterable": false, "sortable": true }, { "key": "cpu", "label": "CPU support", "type": "string", "description": "Processor families or installed processors the named model supports.", "filterable": false, "sortable": true }, { "key": "cpu_sockets", "label": "CPU sockets", "type": "number", "unit": "sockets", "description": "Processors per server.", "filterable": false, "sortable": true }, { "key": "rack_cpu_count", "label": "CPUs per rack", "type": "number", "unit": "CPUs", "description": "Total CPUs in a rack-scale configuration.", "filterable": false, "sortable": true }, { "key": "gpu_memory_gb", "label": "Memory per GPU", "type": "number", "unit": "GB", "description": "Memory on each accelerator in the cited configuration.", "filterable": false, "sortable": true }, { "key": "total_gpu_memory_gb", "label": "Total GPU memory", "type": "number", "unit": "GB", "description": "Vendor-stated total accelerator memory per system.", "filterable": false, "sortable": true }, { "key": "max_system_memory_tb", "label": "Max system memory", "type": "number", "unit": "TB", "description": "Vendor-stated maximum host RAM, not GPU memory.", "filterable": true, "sortable": true }, { "key": "memory_slots", "label": "Memory slots", "type": "number", "unit": "DIMMs", "description": "Host memory DIMM slots.", "filterable": false, "sortable": true }, { "key": "gpu_tdp_w", "label": "GPU TDP", "type": "number", "unit": "W", "description": "Thermal design power per GPU, not measured server draw.", "filterable": false, "sortable": true }, { "key": "system_power_kw", "label": "System power", "type": "number", "unit": "kW", "description": "Maximum power the vendor states for one system, as printed on its own page.", "filterable": true, "sortable": true }, { "key": "rack_power_kw", "label": "Rack power", "type": "number", "unit": "kW", "description": "Power the vendor states for a full rack-scale system.", "filterable": true, "sortable": true }, { "key": "power_supply", "label": "Power supplies", "type": "string", "description": "PSU ratings, quantity and redundancy where published. Rating, not draw.", "filterable": false, "sortable": true }, { "key": "rack_power_delivery", "label": "Rack power delivery", "type": "string", "description": "Power-shelf rating in rack-scale and rack-sled systems. Delivery capacity, not rack demand.", "filterable": false, "sortable": true }, { "key": "interconnect", "label": "GPU interconnect", "type": "string", "description": "GPU fabric and published link bandwidth.", "filterable": false, "sortable": true }, { "key": "networking", "label": "Networking", "type": "string", "description": "Published ports, adapters and speeds.", "filterable": false, "sortable": true }, { "key": "document_date", "label": "Source date", "type": "string", "description": "Publication date printed on the cited page, where the vendor prints one.", "filterable": false, "sortable": true }, { "key": "warranty", "label": "Warranty", "type": "string", "description": "Base warranty or service statement the vendor publishes.", "filterable": false, "sortable": true }, { "key": "operating_temperature_c", "label": "Operating temperature", "type": "string", "unit": "°C", "description": "Supported ambient range with the vendor’s caveats.", "filterable": false, "sortable": true }, { "key": "headquarters", "label": "Maker headquarters", "type": "string", "description": "Headquarters or principal corporate office of the maker, not a factory location.", "filterable": false, "sortable": true } ], "records": [ { "id": "amax-acelemax-axg-428ilc2", "name": "AMAX AceleMax AXG-428ILC2", "url": "https://www.amax.com/product/axg-428ilc2/", "summary": "A liquid-cooled HGX B200 server for AI training, inference, and HPC deployments. CPU and GPU cold plates include leak detection.", "hq": "US", "founded": 1979, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "cooling": "CPU and GPU direct liquid cooling with leak detection", "gpu_count": 8, "power_supply": "8 × 3000 W; 4+4 redundancy; 80 PLUS Titanium", "rack_units": 4, "gpu_model": "NVIDIA HGX B200", "memory_slots": 24, "maker": "AMAX", "headquarters": "Fremont, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://www.amax.com/product/axg-428ilc2/", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.amax.com/about-us/", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "asrock-rack-4u8x-gnr2-dlc-b200", "name": "ASRock Rack 4U8X-GNR2/DLC SYN B200", "url": "https://www.asrockrack.com/general/productdetail.asp?Model=4U8X-GNR2/DLC+SYN+B200", "summary": "A compact HGX B200 platform for AI infrastructure buyers and system integrators. PCIe switch synthetic mode supports GPU, network, and storage data movement.", "hq": "TW", "founded": 2013, "tags": [ "NVIDIA GPUs" ], "fields": { "cpu_sockets": 2, "gpu_count": 8, "power_supply": "8 × 3000 W; 4+4 redundancy; 80 PLUS Titanium", "rack_units": 4, "gpu_model": "NVIDIA HGX B200", "memory_slots": 32, "maker": "ASRock Rack", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://www.asrockrack.com/general/productdetail.asp?Model=4U8X-GNR2/DLC+SYN+B200", "accessed": "2026-11-11", "fields": [ "cpu_sockets", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.asrockrack.com/general/contact.asp", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.asrockrack.com/general/about.asp", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "asus-xa-gb721-e2", "name": "ASUS XA GB721-E2", "url": "https://servers.asus.com/products/asus-ai-pod/nvidia-grace-blackwell/XA-GB721-E2", "summary": "ASUS's liquid-cooled GB300 NVL72 rack-scale solution in a 48RU NVIDIA MGX rack, part of the ASUS AI POD line. For AI cloud and data center buyers who want ASUS rack deployment tooling.", "hq": "TW", "founded": 1989, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "ASUS", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (GB300 NVL72)", "gpu_count": 72, "rack_cpu_count": 36, "cpu": "36 x NVIDIA Grace CPU", "rack_units": 48, "form_factor": "Rack-scale, 1 x NVIDIA MGX 48RU rack, 2236 x 600 x 1159 mm", "cooling_type": "Liquid", "cooling": "Liquid cooling system; supports liquid-to-air and liquid-to-liquid options", "rack_power_delivery": "6 or 8 x 33 kW power shelves, 1 x 1400A rack busbar", "interconnect": "9 NVLink switch trays, 14.4 TB/s switching capacity", "networking": "4 x ConnectX-8 800GbE OSFP ports and 1 x BlueField-3 DPU (B3240) per compute tray", "document_date": "2025-09-04" }, "sources": [ { "url": "https://servers.asus.com/products/asus-ai-pod/nvidia-grace-blackwell/XA-GB721-E2", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor" ] }, { "url": "https://dlcdnet.asus.com/pub/ASUS/server/XA GB721-E2/Datasheet/DataSheet_nvidia-gb300-nvl72_XA_GB721-E2_20250904.pdf", "accessed": "2026-11-11", "fields": [ "cpu", "document_date", "form_factor", "interconnect", "networking", "rack_cpu_count", "rack_power_delivery", "rack_units" ] }, { "url": "https://rog.asus.com/sg/support/faq/1057460/", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "asus-xa-nb3i-e12", "name": "ASUS XA NB3I-E12", "url": "https://servers.asus.com/products/servers/gpu-servers/XA-NB3I-E12", "summary": "An HGX B300 server for demanding AI workloads. Eight embedded ConnectX-8 interfaces support GPU networking at up to 800G per SXM.", "hq": "TW", "founded": 1989, "tags": [ "NVIDIA GPUs" ], "fields": { "networking": "8 embedded ConnectX-8 InfiniBand; XDR/800G per SXM", "gpu_count": 8, "rack_units": 9, "gpu_model": "NVIDIA HGX B300", "memory_slots": 32, "maker": "ASUS", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://servers.asus.com/products/servers/gpu-servers/XA-NB3I-E12", "accessed": "2026-11-11", "fields": [ "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "networking", "rack_units" ] }, { "url": "https://rog.asus.com/sg/support/faq/1057460/", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://press.asus.com/blog/asus-empowers-future-commitment-continuous-innovation/", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "bull-bullsequana-ai-641", "name": "Bull (formerly Eviden) BullSequana AI 641-A02/H02", "url": "https://www.bull.com/en/products/ai/gpu-servers", "summary": "An eight-GPU HGX H200 platform for large-model training and inference. Bull offers air or hybrid cooling configurations in this family.", "hq": "FR", "founded": 1931, "tags": [ "NVIDIA GPUs", "Air or liquid" ], "fields": { "cooling": "Air or hybrid, according to configuration", "gpu_count": 8, "gpu_model": "NVIDIA HGX H200", "cpu": "Intel Xeon", "maker": "Bull (Eviden)", "headquarters": "Les Clayes-sous-Bois, France", "gpu_vendor": "NVIDIA", "cooling_type": "Air or liquid" }, "sources": [ { "url": "https://www.bull.com/en/products/ai/gpu-servers", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_count", "gpu_model", "gpu_vendor" ] }, { "url": "https://www.bull.com/en/about-bull", "accessed": "2026-11-11", "fields": [] }, { "url": "https://www.bull.com/fr/about/histoire-bull", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "cisco-ucs-c880a-m8", "name": "Cisco UCS C880A M8", "url": "https://www.cisco.com/c/en/us/products/collateral/servers-unified-computing/ucs-c-series-rack-servers/ucs-c880a-m8-rack-server-ds.html", "summary": "An air-cooled B300 platform for enterprise and service-provider AI training, fine-tuning, and inference. It integrates one ConnectX-8 adapter per GPU for cluster expansion.", "hq": "US", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "networking": "8 integrated ConnectX-8 adapters for east-west traffic", "cooling": "Air", "gpu_count": 8, "rack_units": 10, "document_date": "2026-08-13", "gpu_model": "NVIDIA HGX B300 NVL8", "memory_slots": 32, "maker": "Cisco", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://www.cisco.com/c/en/us/products/collateral/servers-unified-computing/ucs-c-series-rack-servers/ucs-c880a-m8-rack-server-ds.html", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "networking", "rack_units" ] }, { "url": "https://www.cisco.com/site/us/en/about/contact-cisco/index.html", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.cisco.com/c/dam/en_us/about/ac49/ac55/docs/Catalyst_Report.pdf", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "colfax-cx7850q-xk9", "name": "Colfax CX7850q-XK9", "url": "https://www.colfax-intl.com/Servers/CX7850q-XK9", "summary": "A 7U HGX H200 server for AI and HPC teams, built to order. It carries eight NVLink-connected H200 GPUs, eighteen front NVMe bays and 3+3 redundant 4000 W power supplies.", "hq": "US", "founded": 1987, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "gpu_count": 8, "power_supply": "6 × 4000 W; 3+3 redundancy", "rack_units": 7, "gpu_model": "NVIDIA HGX H200 8-GPU SXM5", "max_system_memory_tb": 8, "memory_slots": 32, "maker": "Colfax", "headquarters": "Santa Clara, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air", "cpu": "4th/5th Gen Intel Xeon Scalable, up to 350 W TDP", "cpu_sockets": 2, "cooling": "8 hot-swap dual-rotor fans (N+1)" }, "sources": [ { "url": "https://www.colfax-intl.com/Servers/CX7850q-XK9", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.colfax-intl.com/about/aboutus", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "compal-sg720-2a", "name": "Compal SG720-2A", "url": "https://www.storagenewsletter.com/2025/06/16/amd-advancing-ai-isc-2025-compal-optimizes-ai-workloads-with-amd-instinct-mi355x/", "summary": "An eight-GPU AMD Instinct MI350 Series server from Compal, offered with air, single-phase or two-phase liquid cooling, with an ORv3 21-inch version (OG720-2A).", "hq": "TW", "tags": [ "AMD GPUs", "Air or liquid" ], "fields": { "maker": "Compal", "gpu_vendor": "AMD", "gpu_model": "AMD Instinct MI355X (one supported option)", "gpu_count": 8, "gpu_memory_gb": 288, "cooling_type": "Air or liquid", "cooling": "Air cooling; single-phase and two-phase direct liquid cooling (two-phase with ZutaCore HyperCool)", "interconnect": "PCIe Gen5; AMD Infinity Fabric" }, "sources": [ { "url": "https://www.storagenewsletter.com/2025/06/16/amd-advancing-ai-isc-2025-compal-optimizes-ai-workloads-with-amd-instinct-mi355x/", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "interconnect" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9680", "name": "Dell PowerEdge XE9680", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9680", "summary": "An eight-accelerator platform for enterprise AI training, inference, and HPC. This record captures the H200 GPU option and the server’s air-cooled rack format.", "hq": "US", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air", "gpu_memory_gb": 141, "gpu_tdp_w": 700, "gpu_count": 8, "rack_units": 6, "gpu_model": "NVIDIA HGX H200 SXM5 (one supported option)", "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://www.dell.com/support/manuals/en-au/poweredge-xe9680/xe9680_ism_pub/gpu-specifications?guid=guid-1f91b9fb-b8fd-492e-863f-45febc9355c6&lang=en-us", "accessed": "2026-11-11", "fields": [ "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor" ] }, { "url": "https://www.dell.com/en-us/infohub/articles/poweredge-xe9680-rack-integration", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "rack_units" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9712", "name": "Dell PowerEdge XE9712", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9712", "summary": "Dell's 1RU Grace Blackwell Ultra compute sled for GB300 NVL72, shipped in the 48RU IR9048 rack with 33 kW power shelves. Aimed at large-scale LLM training and real-time inference buyers who want a Dell-integrated rack.", "hq": "US", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air or liquid", "Rack-scale" ], "fields": { "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra", "cpu": "NVIDIA Grace, 72 cores per processor", "cooling_type": "Air or liquid", "cooling": "Air cooling and Direct Liquid Cooling (DLC)", "form_factor": "1 RU sled in a 48 RU IR9048 rack", "gpu_memory_gb": 288, "rack_power_delivery": "Power shelf 33 kW (PS33), 6 x 5500 W AC PSUs, feeding the sled through the IR9048 busbar", "interconnect": "NVLink CPU-GPU interconnect with 900 GB/s coherent memory" }, "sources": [ { "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9712", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "form_factor", "gpu_memory_gb", "gpu_model", "gpu_vendor", "interconnect", "rack_power_delivery" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9780", "name": "Dell PowerEdge XE9780", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9780", "summary": "An air-cooled Intel Xeon 6 server for large-model training and inference. Dell lists both B200 and B300 accelerator configurations.", "hq": "US", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air", "gpu_count": 8, "power_supply": "12 × 3200 W Titanium", "gpu_model": "NVIDIA HGX B300 NVL8 (one supported option)", "max_system_memory_tb": 4, "memory_slots": 32, "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9780", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "power_supply" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9780l", "name": "Dell PowerEdge XE9780L", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9780l", "summary": "Dell's liquid-cooled 3 OU node with eight HGX B300 GPUs for ORv3 racks and a 33 kW power shelf. Built for AI factories that run IR7000 rack-scale deployments.", "hq": "US", "founded": 1984, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300 (one supported option)", "gpu_count": 8, "gpu_tdp_w": 1100, "form_factor": "3 OU ORv3 node", "cooling_type": "Liquid", "cooling": "Liquid-cooled CPUs, GPUs and NVLink switches; 4 high-performance fans for the UBB and 8 for the HPM", "cpu": "2 × Intel Xeon 6", "cpu_sockets": 2, "memory_slots": 32, "rack_power_delivery": "33 kW power shelf with 6 × 5500 W AC PSUs" }, "sources": [ { "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9780l", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_model", "gpu_tdp_w", "gpu_vendor", "memory_slots", "rack_power_delivery" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9785", "name": "Dell PowerEdge XE9785", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9785", "summary": "Dell's air-cooled 10U eight-GPU server on dual 5th Gen EPYC, with AMD Instinct MI355X accelerators. For AI training and inference without a liquid-cooling build-out.", "hq": "US", "founded": 1984, "tags": [ "AMD GPUs", "Air-cooled" ], "fields": { "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "AMD", "gpu_model": "AMD Instinct MI355X (one supported option)", "gpu_count": 8, "gpu_memory_gb": 288, "gpu_tdp_w": 1400, "rack_units": 10, "cooling_type": "Air", "cooling": "Advanced air-cooled", "cpu": "Two 5th Gen AMD EPYC 9005 Series, up to 192 cores each", "cpu_sockets": 2, "max_system_memory_tb": 6, "memory_slots": 24, "power_supply": "12 × 3200 W Titanium; 200-240 VAC; hot-swap redundant", "interconnect": "AMD Infinity Fabric (MI355X)" }, "sources": [ { "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9785", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor", "interconnect", "max_system_memory_tb", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "dell-poweredge-xe9785l", "name": "Dell PowerEdge XE9785L", "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9785l", "summary": "A liquid-cooled AMD EPYC rack sled for dense enterprise and service-provider AI deployments. The server supports AMD MI355X and NVIDIA B300 accelerator configurations.", "hq": "US", "founded": 1984, "tags": [ "AMD GPUs", "Liquid-cooled" ], "fields": { "cooling": "Direct-to-chip liquid cooling for CPUs and GPUs", "rack_power_delivery": "33 kW power shelf", "gpu_count": 8, "gpu_model": "AMD Instinct MI355X (one supported option)", "memory_slots": 24, "form_factor": "3 Open Rack Units per compute node", "maker": "Dell Technologies", "headquarters": "Round Rock, United States", "gpu_vendor": "AMD", "cooling_type": "Liquid" }, "sources": [ { "url": "https://www.dell.com/en-us/shop/ipovw/poweredge-xe9785l", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "rack_power_delivery" ] }, { "url": "https://i.dell.com/sites/csdocuments/Corporate_financials_Documents/en/us/10k-fy05.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "exxact-tensorex-ts4-128275176", "name": "Exxact TensorEX TS4-128275176", "url": "https://www.exxactcorp.com/Exxact-TS4-128275176-E128275176", "summary": "A configurable HGX B200 system for AI development, research, and simulation. Exxact offers deployment with its machine-learning software environments.", "hq": "US", "founded": 1992, "tags": [ "NVIDIA GPUs" ], "fields": { "interconnect": "NVLINK; up to 1.8 TB/s", "gpu_count": 8, "rack_units": 8, "gpu_model": "NVIDIA B200 SXM", "memory_slots": 32, "cpu": "4th/5th Gen Intel Xeon Scalable", "maker": "Exxact", "headquarters": "Fremont, United States", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://www.exxactcorp.com/Exxact-TS4-128275176-E128275176", "accessed": "2026-11-11", "fields": [ "cpu", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "rack_units" ] }, { "url": "https://www.exxactcorp.com/blog/News/take-a-first-look-inside-the-new-exxact-headquarters", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "fsas-primergy-gx2570-m8s", "name": "Fsas Technologies (Fujitsu) PRIMERGY GX2570 M8s", "url": "https://jp.fujitsu.com/platform/server/primergy/products/lineup/gx2570m8s/", "summary": "A B200 GPU server for AI and high-performance computing buyers. Fsas Technologies offers both air-cooled and water-cooled chassis configurations.", "hq": "JP", "founded": 1989, "tags": [ "NVIDIA GPUs", "Air or liquid" ], "fields": { "cooling": "Air or water cooling", "gpu_count": 8, "gpu_model": "NVIDIA HGX B200", "max_system_memory_tb": 6, "memory_slots": 24, "form_factor": "10U air-cooled; 4U water-cooled", "maker": "Fsas Technologies (Fujitsu)", "headquarters": "Kawasaki, Japan", "gpu_vendor": "NVIDIA", "cooling_type": "Air or liquid" }, "sources": [ { "url": "https://jp.fujitsu.com/platform/server/primergy/products/lineup/gx2570m8s/gx2570m8s_catalog.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots" ] }, { "url": "https://info.archives.global.fujitsu/jp/news/2023/12/en/26.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g4l3-sd1-lax5", "name": "GIGABYTE G4L3-SD1-LAX5", "url": "https://www.gigabyte.com/Enterprise/GPU-Server/G4L3-SD1-LAX5", "summary": "4U dual-Xeon server with a direct-liquid-cooled NVIDIA HGX B200 8-GPU baseboard. Built for high-density B200 racks that run on facility liquid cooling.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B200", "gpu_count": 8, "rack_units": 4, "form_factor": "4U rackmount", "cooling_type": "Liquid", "cooling": "CPU+GPU direct liquid cooling with leak detection", "cpu": "Dual 5th/4th Gen Intel Xeon Scalable or Intel Xeon CPU Max Series, up to 350 W", "cpu_sockets": 2, "memory_slots": 32, "power_supply": "4+4 × 3000 W 80 PLUS Titanium redundant", "interconnect": "1.8 TB/s GPU-to-GPU via NVLink and NVSwitch", "document_date": "2026-08-27", "operating_temperature_c": "10 to 35", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G4L3-SD1-LAX5_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "document_date", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g4l4-sd3-lax7", "name": "GIGABYTE G4L4-SD3-LAX7", "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G4L4-SD3-LAX7_datasheet.pdf", "summary": "Gigabyte's 4U liquid-cooled HGX B300 server with an integrated-manifold DLC design and leak detection. Built for dense liquid-cooled B300 racks.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "interconnect": "NVLink with NVSwitch, 1.8 TB/s GPU-to-GPU", "document_date": "2026-07-20", "rack_units": 4, "cooling_type": "Liquid", "cooling": "Integrated-manifold DLC solution with leak detection", "cpu": "2 × Intel Xeon 6700/6500-Series", "cpu_sockets": 2, "memory_slots": 32, "power_supply": "5+5 × 3000 W 80 PLUS Titanium redundant", "networking": "8 × 800 Gb/s OSFP ConnectX-8 SuperNIC (InfiniBand XDR or dual 400 Gb/s Ethernet)", "operating_temperature_c": "10-35", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G4L4-SD3-LAX7_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g893-sd1-aax5", "name": "GIGABYTE G893-SD1-AAX5", "url": "https://www.gigabyte.com/Enterprise/GPU-Server/G893-SD1-AAX5", "summary": "An Intel-based HGX B200 server for AI training, inference, and HPC. It combines NVLink GPU connectivity with PCIe Gen5 expansion.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs" ], "fields": { "interconnect": "NVLink and NVSwitch; 1.8 TB/s GPU-to-GPU bandwidth", "power_supply": "12 × 3000 W; 6+6 redundancy; 80 PLUS Titanium", "rack_units": 8, "gpu_model": "NVIDIA HGX B200", "memory_slots": 32, "maker": "GIGABYTE", "headquarters": "New Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://www.gigabyte.com/Enterprise/GPU-Server/G893-SD1-AAX5", "accessed": "2026-11-11", "fields": [ "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g893-sg1-aax1", "name": "GIGABYTE G893-SG1-AAX1", "url": "https://www.gigabyte.com/Enterprise/GPU-Server/G893-SG1-AAX1", "summary": "8U dual-Xeon server carrying eight Intel Gaudi 3 HL-325L OAM cards on an HLB-325 baseboard. Aimed at enterprise Gaudi 3 AI training and inference clusters.", "hq": "TW", "founded": 1986, "tags": [ "Intel Gaudi" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "Intel", "gpu_model": "Intel Gaudi 3 HL-325L OAM", "gpu_count": 8, "rack_units": 8, "form_factor": "8U rackmount", "cpu": "Dual 5th/4th Gen Intel Xeon Scalable or Intel Xeon CPU Max Series, up to 350 W", "cpu_sockets": 2, "memory_slots": 32, "power_supply": "5+5 × 3000 W 80 PLUS Titanium redundant", "networking": "6 × 800 Gb/s OSFP (RoCE v2), 24 × 200 Gb/s Ethernet for accelerator scale-out", "document_date": "2026-01-02", "operating_temperature_c": "10 to 30", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G893-SG1-AAX1_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g894-ad3-aax7", "name": "GIGABYTE G894-AD3-AAX7", "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-AD3-AAX7_datasheet.pdf", "summary": "Gigabyte's 8U HGX B300 server with dual Xeon 6900 and 24 DIMM slots. Built for memory-heavy AI and HPC clusters.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "interconnect": "NVLink with NVSwitch, 1.8 TB/s GPU-to-GPU", "document_date": "2026-06-22", "rack_units": 8, "cpu": "2 × Intel Xeon 6900E+/6900-Series", "cpu_sockets": 2, "memory_slots": 24, "power_supply": "6+6 × 3000 W 80 PLUS Titanium redundant", "networking": "8 × 800 Gb/s OSFP ConnectX-8 SuperNIC (InfiniBand XDR or dual 400 Gb/s Ethernet)", "operating_temperature_c": "10-30", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-AD3-AAX7_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g894-sd3-aax7", "name": "GIGABYTE G894-SD3-AAX7", "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-SD3-AAX7_datasheet.pdf", "summary": "Gigabyte's 8U HGX B300 server with dual Xeon 6700/6500 and 32 DIMM slots. Built for AI training and inference clusters.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "interconnect": "NVLink with NVSwitch, 1.8 TB/s GPU-to-GPU", "document_date": "2026-08-24", "rack_units": 8, "cpu": "2 × Intel Xeon 6700/6500-Series", "cpu_sockets": 2, "memory_slots": 32, "power_supply": "12 × 3000 W 80 PLUS Titanium redundant", "networking": "8 × 800 Gb/s OSFP ConnectX-8 SuperNIC (InfiniBand XDR or dual 400 Gb/s Ethernet)", "operating_temperature_c": "10-30", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-SD3-AAX7_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-g894-zd3-aax7", "name": "GIGABYTE G894-ZD3-AAX7", "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-ZD3-AAX7_datasheet.pdf", "summary": "Gigabyte's 8U HGX B300 server with dual AMD EPYC 9005/9004 and 24 DIMM slots. Built for AI clusters that standardize on EPYC hosts.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs" ], "fields": { "maker": "GIGABYTE", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "interconnect": "NVLink with NVSwitch, 1.8 TB/s GPU-to-GPU", "document_date": "2026-04-21", "rack_units": 8, "cpu": "2 × AMD EPYC 9005/9004 Series", "cpu_sockets": 2, "memory_slots": 24, "power_supply": "12 × 3000 W 80 PLUS Titanium redundant", "networking": "8 × 800 Gb/s OSFP ConnectX-8 SuperNIC (InfiniBand XDR or dual 400 Gb/s Ethernet)", "operating_temperature_c": "10-30", "headquarters": "New Taipei, Taiwan" }, "sources": [ { "url": "https://download.gigabyte.com/FileList/DataSheet/GIGABYTE%20G894-ZD3-AAX7_datasheet.pdf", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gigabyte-gigapod-gb300-nvl72", "name": "GIGABYTE GIGAPOD NVIDIA GB300 NVL72", "url": "https://www.gigabyte.com/Enterprise/GIGAPOD-Rack-Scale/AI-DLC-Rack_NVIDIA-GB300-NVL72", "summary": "GIGABYTE's direct-liquid-cooled GB300 NVL72 rack built from 18 XN15-CB0-L01 compute trays and 9 NVLink switch trays, also sold as 16+8 rack GIGAPOD scalable units. For AI factory builders who scale from one rack to a full pod.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "GIGABYTE", "headquarters": "New Taipei, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (GB300 NVL72)", "gpu_count": 72, "rack_cpu_count": 36, "cpu": "36 x NVIDIA Grace CPU", "rack_units": 48, "cooling_type": "Liquid", "cooling": "Direct liquid cooling; in-rack liquid-to-liquid CDU (200 kW) or in-row CDU (1350 kW)", "rack_power_kw": 140, "rack_power_delivery": "6 x 1U 33 kW power shelves", "networking": "4 x 800 Gb/s OSFP ConnectX-8 SuperNIC and 1 BlueField-3 DPU per compute tray" }, "sources": [ { "url": "https://www.gigabyte.com/Enterprise/GIGAPOD-Rack-Scale/AI-DLC-Rack_NVIDIA-GB300-NVL72", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_count", "gpu_model", "gpu_vendor", "networking", "rack_cpu_count", "rack_power_delivery", "rack_power_kw", "rack_units" ] }, { "url": "https://www.gigabyte.com/FileUpload/Global/SiteMap/77/images/2015GIGABYTE_report_2016.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "h3c-uniserver-r5500-g7-x", "name": "H3C UniServer R5500 G7 X", "url": "https://www.h3c.com/en/Products_and_Solutions/Computing/UniServer_Series/Products_With_Picture/GPU/R5500_G7_X/R5500_G7_X/", "summary": "A modular eight-GPU server for distributed model training and enterprise AI workloads. GPU and CPU modules support separate power planes and servicing.", "hq": "CN", "founded": 2003, "tags": [], "fields": { "interconnect": "Full mesh GPU interconnect", "cpu_sockets": 2, "gpu_count": 8, "rack_units": 10, "max_system_memory_tb": 8, "memory_slots": 32, "maker": "H3C", "headquarters": "Beijing, China", "gpu_vendor": null, "cooling_type": null }, "sources": [ { "url": "https://www.h3c.com/en/Products_and_Solutions/Computing/UniServer_Series/Products_With_Picture/GPU/R5500_G7_X/R5500_G7_X/", "accessed": "2026-11-11", "fields": [ "cpu_sockets", "gpu_count", "interconnect", "max_system_memory_tb", "memory_slots", "rack_units" ] }, { "url": "https://www.h3c.com/en/About_Us/Contact_Us/Headquarters/", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.h3c.com/cn/Aspx/about_h3C/profile.html", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "hpe-compute-xd690", "name": "HPE Compute XD690", "url": "https://buy.hpe.com/us/en/compute/rack-servers/proliant-compute-xd600-servers/hpe-compute-xd690/p/1014909740", "summary": "HPE's 10U air-cooled server with eight Blackwell Ultra B300 GPUs and two Xeon 6 processors. Built for large AI training, tuning and inference clusters.", "hq": "US", "founded": 2015, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "maker": "HPE", "headquarters": "Spring, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 10, "cooling_type": "Air", "cpu": "2 × Intel Xeon 6", "cpu_sockets": 2, "networking": "HPE InfiniBand XDR and NVIDIA Spectrum-X Ethernet support" }, "sources": [ { "url": "https://buy.hpe.com/us/en/compute/rack-servers/proliant-compute-xd600-servers/hpe-compute-xd690/p/1014909740", "accessed": "2026-11-11", "fields": [ "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_model", "gpu_vendor", "networking", "rack_units" ] }, { "url": "https://www.hpe.com/int/en/about/enterprise-technology-us.html", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "hpe-proliant-compute-xd685", "name": "HPE ProLiant Compute XD685", "url": "https://www.hpe.com/int/en/collaterals/collateral.a00073553enw.html", "summary": "An AMD EPYC platform for AI service providers, large enterprises, and sovereign AI deployments. Chassis height depends on the air or direct liquid cooling configuration.", "hq": "US", "founded": 2015, "tags": [ "NVIDIA GPUs", "Air or liquid" ], "fields": { "cooling": "Air or direct liquid cooling", "gpu_count": 8, "gpu_model": "NVIDIA B300 (direct-liquid-cooled option)", "form_factor": "5U direct liquid cooling; 6U air cooling", "cpu": "5th Generation AMD EPYC", "maker": "HPE", "headquarters": "Spring, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air or liquid" }, "sources": [ { "url": "https://www.hpe.com/int/en/collaterals/collateral.a00073553enw.html", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "form_factor", "gpu_count", "gpu_model", "gpu_vendor" ] }, { "url": "https://www.hpe.com/int/en/about/enterprise-technology-us.html", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "huawei-atlas-800t-a2", "name": "Huawei Atlas 800T A2", "url": "https://e.huawei.com/cn/products/computing/ascend/atlas-800t-a2", "summary": "An AI training server for government, finance, education, research, and telecom deployments. It uses Kunpeng CPUs and provides eight 200GE network interfaces.", "hq": "CN", "founded": 1987, "tags": [ "Air-cooled" ], "fields": { "cpu_sockets": 4, "networking": "8 × 200GE QSFP; RoCE", "cooling": "Air", "power_supply": "4 × 2.6 kW hot-swap modules; 2+2 redundancy", "rack_units": 4, "memory_slots": 32, "cpu": "Huawei Kunpeng 920", "maker": "Huawei", "headquarters": "Shenzhen, China", "gpu_vendor": null, "cooling_type": "Air" }, "sources": [ { "url": "https://e.huawei.com/cn/products/computing/ascend/atlas-800t-a2", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "memory_slots", "networking", "power_supply", "rack_units" ] }, { "url": "https://www.huawei.com/en/media-center/company-facts", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "hypertec-ciara-titan-gs685r-g6-b200", "name": "Hypertec CIARA TITAN GS685R-G6 B200", "url": "https://hypertec.com/products/ai-gpu-servers/titan-gs685r-g6-b200", "summary": "An AMD EPYC HGX B200 platform for enterprise-scale model training, inference, and scientific computing. It uses passive GPU air cooling and redundant power.", "hq": "CA", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Passive air cooling", "gpu_count": 8, "power_supply": "12 × 3000 W; 6+6 redundancy; 80 PLUS Titanium", "rack_units": 8, "gpu_model": "NVIDIA HGX B200", "max_system_memory_tb": 6, "maker": "Hypertec", "headquarters": "Montreal, Canada", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://hypertec.com/products/ai-gpu-servers/titan-gs685r-g6-b200", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "power_supply", "rack_units" ] }, { "url": "https://hypertec.com/news/hypertec-an-international-provider-of-innovative-information-technology-solutions-named-finalist-in-les-mercuriades", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "ieit-inspur-nf5688g7-amd", "name": "IEIT SYSTEMS (Inspur) NF5688G7 AMD", "url": "https://en.ieisystem.com/product/ai/10969.html", "summary": "A hyperscale AI training server for on-premises and cloud deployments. The AMD configuration combines eight Hopper accelerators with EPYC 9004 processors.", "hq": "CN", "founded": 1998, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air; NF5688-A7-A0-R0-00 configuration", "operating_temperature_c": "10–35", "gpu_tdp_w": 700, "gpu_count": 8, "rack_units": 6, "gpu_model": "NVIDIA HGX Hopper", "cpu": "AMD EPYC 9004", "maker": "IEIT Systems (Inspur)", "headquarters": "Jinan, China", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://en.ieisystem.com/product/ai/10969.html", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_count", "gpu_model", "gpu_tdp_w", "gpu_vendor", "operating_temperature_c", "rack_units" ] }, { "url": "https://en.ieisystem.com/about/news/9171.html", "accessed": "2026-11-11", "fields": [] }, { "url": "https://static.cninfo.com.cn/finalpage/enpage/000977_2.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "ingrasys-nvidia-gb300-nvl72", "name": "Ingrasys (Foxconn) NVIDIA GB300 NVL72", "url": "https://www.ingrasys.com/solutions/NVIDIA/nvidia_gb300_nvl72/", "summary": "A liquid-cooled rack-scale system for AI reasoning, inference, and training. Ingrasys integrates Grace CPUs, Blackwell Ultra GPUs, and NVLink networking.", "hq": "TW", "founded": 2002, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "cooling": "Liquid", "rack_power_delivery": "33 kW maximum output per power shelf", "gpu_count": 72, "gpu_model": "NVIDIA Blackwell Ultra", "form_factor": "Rack-scale", "rack_cpu_count": 36, "maker": "Ingrasys (Foxconn)", "headquarters": "Taoyuan, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://www.ingrasys.com/solutions/NVIDIA/nvidia_gb300_nvl72/", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "rack_cpu_count", "rack_power_delivery" ] }, { "url": "https://www.ingrasys.com/solutions/NVIDIA/nvidia_overview/", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.ingrasys.com/company/about/", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "inventec-artemis-ii-gb300-nvl72", "name": "Inventec Artemis II Rack (GB300 NVL72)", "url": "https://ebg.inventec.com/en/solution/NVIDIA/Artemis%20II%20Rack", "summary": "Inventec's liquid-cooled rack for NVIDIA GB300 NVL72 building blocks with sidecar or row-based cooling and leak detection. For hyperscalers and cloud builders sourcing ODM racks.", "hq": "TW", "founded": 1975, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "Inventec", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA GB300 NVL72", "cooling_type": "Liquid", "cooling": "Liquid cooling in sidecar or row-based configurations with leakage detection", "rack_power_kw": 136, "interconnect": "NVIDIA NVLink cable cartridge and manifold, blind-mate design" }, "sources": [ { "url": "https://ebg.inventec.com/en/solution/NVIDIA/Artemis%20II%20Rack", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_model", "gpu_vendor", "interconnect", "rack_power_kw" ] }, { "url": "https://www.inventec.com/storage/upload/csr4/file/2023-07-26/F41EKMWZCcoMJVvgenuPk33ORgBpdBTBD7v5BeuL.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "inventec-p9000ig6-ac", "name": "Inventec P9000IG6(AC)", "url": "https://ebg.inventec.com/en/product/Server/10U/P9000IG6%28AC%29", "summary": "An air-cooled HGX B200 platform from Inventec’s server business. It targets AI training, scientific computing, and other data-intensive enterprise workloads.", "hq": "TW", "founded": 1975, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air", "gpu_count": 8, "power_supply": "54 V subsystem: 3300 W PSUs, 4+1 or 4+2 redundancy", "rack_units": 10, "gpu_model": "NVIDIA HGX B200", "maker": "Inventec", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://ebg.inventec.com/en/product/Server/10U/P9000IG6%28AC%29", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor", "power_supply", "rack_units" ] }, { "url": "https://www.inventec.com/storage/upload/csr4/file/2023-07-26/F41EKMWZCcoMJVvgenuPk33ORgBpdBTBD7v5BeuL.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.inventec.com/en/", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "lenovo-nvidia-gb300-nvl72", "name": "Lenovo NVIDIA GB300 NVL72", "url": "https://lenovopress.lenovo.com/datasheet/ds0207-lenovo-nvidia-gb300-nvl72", "summary": "Lenovo's 48U MGX rack with 18 compute trays and 9 NVLink switch trays, cooled by warm-water hybrid liquid and air. For enterprises and cloud providers that want Lenovo services and warm-water efficiency.", "hq": "CN", "founded": 1984, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "Lenovo", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (B300)", "gpu_count": 72, "rack_units": 48, "form_factor": "Rack-scale, 48U MGX rack with 18 x 1U compute trays and 9 x 1U NVLink switch trays", "cooling_type": "Liquid", "cooling": "Hybrid liquid and air cooling with PG25 or DI water, warm water operation 18-45 C", "rack_power_delivery": "Up to eight 33 kW power shelves and a 48V DC busbar", "interconnect": "5th generation NVLink, 130 TB/s GPU-to-GPU bandwidth across the NVLink domain", "networking": "Integrated 800 Gb/s ConnectX-8 networking, dual 25GbE LOM, InfiniBand and Ethernet fabric support", "document_date": "2026-03-16", "headquarters": "Beijing, China; Morrisville, United States" }, "sources": [ { "url": "https://lenovopress.lenovo.com/datasheet/ds0207-lenovo-nvidia-gb300-nvl72", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "document_date", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "networking", "rack_power_delivery", "rack_units" ] }, { "url": "https://investor.lenovo.com/en/sustainability/reports/FY2015-Lenovo-Sustainability-Report.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "lenovo-thinksystem-sc777-v4", "name": "Lenovo ThinkSystem SC777 V4 Neptune", "url": "https://lenovopress.lenovo.com/lp2048-thinksystem-sc777-v4", "summary": "Fanless, fully warm-water-cooled GB200 NVL4 node with four Blackwell GPUs and two Grace CPUs, mounted eight to a 13U N1380 enclosure. For HPC and AI training sites with facility water.", "hq": "CN", "founded": 1984, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "Lenovo", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA GB200 NVL4 (4 × B200 per node)", "gpu_count": 4, "form_factor": "Node in 13U N1380 Neptune enclosure (up to 8 nodes)", "cooling_type": "Liquid", "cooling": "100% direct warm-water cooling, no system fans; inlet water up to 45°C", "cpu": "2 × NVIDIA Grace, 72 Arm Neoverse V2 cores each", "gpu_memory_gb": 186, "total_gpu_memory_gb": 744, "gpu_tdp_w": 1200, "rack_power_delivery": "N1380 enclosure: up to 4 × 15 kW water-cooled power conversion stations (N+1); 18/36/54 kW DC with 2/3/4 PCS", "interconnect": "NVLink 1.8 TB/s GPU full mesh; NVLink-C2C 300 GB/s CPU-GPU", "networking": "2 × 25GbE SFP28 onboard; up to 6 front PCIe 5.0 x16 slots (ConnectX-7/8 or 2 × 800Gb OSFP)", "document_date": "2026-07-15", "headquarters": "Beijing, China; Morrisville, United States" }, "sources": [ { "url": "https://lenovopress.lenovo.com/lp2048-thinksystem-sc777-v4", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "document_date", "form_factor", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor", "interconnect", "networking", "rack_power_delivery", "total_gpu_memory_gb" ] }, { "url": "https://investor.lenovo.com/en/sustainability/reports/FY2015-Lenovo-Sustainability-Report.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "lenovo-thinksystem-sr680a-v3-b200", "name": "Lenovo ThinkSystem SR680a V3 with B200", "url": "https://lenovopress.lenovo.com/lp2247-thinksystem-sr680a-v3-with-b200", "summary": "An Intel-based B200 server for data-center AI training, simulation, and scientific research. The system fits air-cooled facilities.", "hq": "CN", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cpu_sockets": 2, "cooling": "Air", "gpu_memory_gb": 180, "gpu_tdp_w": 1000, "gpu_count": 8, "rack_units": 8, "gpu_model": "NVIDIA B200", "maker": "Lenovo", "headquarters": "Beijing, China; Morrisville, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://lenovopress.lenovo.com/lp2247-thinksystem-sr680a-v3-with-b200", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu_sockets", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor", "rack_units" ] }, { "url": "https://investor.lenovo.com/en/sustainability/reports/FY2015-Lenovo-Sustainability-Report.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "lenovo-thinksystem-sr680a-v4", "name": "Lenovo ThinkSystem SR680a V4", "url": "https://lenovopress.lenovo.com/lp2264-thinksystem-sr680a-v4-server", "summary": "Lenovo's 8U air-cooled server with eight HGX B300 GPUs and two Xeon 6700 processors. Built for enterprise AI training and inference.", "hq": "CN", "founded": 1984, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "maker": "Lenovo", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 8, "cooling_type": "Air", "cooling": "Air-cooled, front-to-rear airflow; 6 × front 60 mm and 15 × rear 80 mm dual-rotor fans, N+1", "cpu": "2 × Intel Xeon 6700-series (P-core)", "cpu_sockets": 2, "gpu_memory_gb": 288, "total_gpu_memory_gb": 2300, "memory_slots": 32, "max_system_memory_tb": 4, "power_supply": "8 × CRPS Premium 80 PLUS Titanium: 3200 W AC (N+1) or 3800 W HVAC/HVDC (N+1 or N+N)", "networking": "8 × 800 Gb/s OSFP ConnectX-8 with GPUDirect; 4 × PCIe 5.0 x16 FHHL; 1 × OCP 3.0", "document_date": "2026-10-05", "headquarters": "Beijing, China; Morrisville, United States" }, "sources": [ { "url": "https://lenovopress.lenovo.com/lp2264-thinksystem-sr680a-v4-server", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "networking", "power_supply", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://investor.lenovo.com/en/sustainability/reports/FY2015-Lenovo-Sustainability-Report.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "lenovo-thinksystem-sr780a-v3", "name": "Lenovo ThinkSystem SR780a V3", "url": "https://lenovopress.lenovo.com/lp1980-thinksystem-sr780a-v3-server", "summary": "A liquid-cooled Intel server for AI training, simulation, and scientific workloads. It supports H100, H200, and B200 GPU configurations.", "hq": "CN", "founded": 1984, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "cooling": "Liquid", "gpu_count": 8, "rack_units": 5, "document_date": "2026-09-18", "gpu_model": "NVIDIA B200 SXM6 (one supported option)", "max_system_memory_tb": 4, "memory_slots": 32, "maker": "Lenovo", "headquarters": "Beijing, China; Morrisville, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://lenovopress.lenovo.com/lp1980-thinksystem-sr780a-v3-server", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "rack_units" ] }, { "url": "https://investor.lenovo.com/en/sustainability/reports/FY2015-Lenovo-Sustainability-Report.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "mitac-olympus-g8825z5u2bc", "name": "MiTAC Olympus G8825Z5U2BC-350X-755", "url": "https://www.mitaccomputing.com/products/detail/G8825Z5U2BC-350X-755", "summary": "An all-AMD server for large language models, AI training and inference, and HPC. The named configuration includes EPYC 9755 processors and MI350X accelerators.", "hq": "TW", "founded": 2014, "tags": [ "AMD GPUs" ], "fields": { "cpu_sockets": 2, "gpu_count": 8, "rack_units": 8, "gpu_model": "AMD Instinct MI350X", "max_system_memory_tb": 6, "memory_slots": 24, "cpu": "AMD EPYC 9755", "maker": "MiTAC", "headquarters": "Taiwan", "gpu_vendor": "AMD", "cooling_type": null }, "sources": [ { "url": "https://www.mitaccomputing.com/products/detail/G8825Z5U2BC-350X-755", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "rack_units" ] }, { "url": "https://www.mitaccomputing.com/en/company/about_mitac", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "msi-cg480-s5063", "name": "MSI CG480-S5063", "url": "https://eps.msi.com/en/product/server/CG480-S5063", "summary": "A 4U NVIDIA MGX platform for enterprise AI training, inference, simulation, and HPC. It supports up to eight dual-width PCIe GPUs.", "hq": "TW", "founded": 1986, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air", "gpu_tdp_w": 600, "gpu_count": 8, "rack_units": 4, "gpu_model": "NVIDIA RTX PRO 6000 Blackwell Server Edition", "max_system_memory_tb": 8, "memory_slots": 32, "maker": "MSI", "headquarters": "New Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://www.msi.com/news/detail/MSI-Expands-NVIDIA-RTX-PRO-Server-Lineup---Accelerated-by-RTX-PRO-6000-Blackwell-Server-Edition-GPU-146930", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_tdp_w", "gpu_vendor", "max_system_memory_tb", "memory_slots", "rack_units" ] }, { "url": "https://storage-asset.msi.com/file/pdf/2025_msi_csr_eng.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "netweb-tyrone-camarero-pda200a2hg-812", "name": "Netweb Tyrone Camarero PDA200A2HG-812", "url": "https://tyronesystems.com/servers/PDA200A2HG-812.php", "summary": "An AMD EPYC HGX B200 server for AI and high-performance computing buyers. The platform includes NVSwitch and twelve hot-swap NVMe bays.", "hq": "IN", "founded": 1999, "tags": [ "NVIDIA GPUs" ], "fields": { "gpu_count": 8, "power_supply": "12 × 3000 W; 6+6 redundancy; Titanium", "rack_units": 8, "gpu_model": "NVIDIA HGX B200", "memory_slots": 24, "cpu": "AMD EPYC 9005/9004", "maker": "Netweb", "headquarters": "Faridabad, India", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://tyronesystems.com/servers/PDA200A2HG-812.php", "accessed": "2026-11-11", "fields": [ "cpu", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.netwebindia.com/investors/intimation-schedule-meet/Intimation_of_investor_Meet_25092026.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.netwebindia.com/investors/board-meeting/2024-25/Q4/Earning_Presentation_q4.pdf", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "nvidia-dgx-b200", "name": "NVIDIA DGX B200", "url": "https://www.nvidia.com/en-us/data-center/dgx-b200/", "summary": "An integrated B200 AI server for organizations developing and running large models. NVIDIA supplies the compute, networking, and system software platform.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs" ], "fields": { "system_power_kw": 14.3, "gpu_count": 8, "power_supply": "6 × 3.3 kW", "rack_units": 10, "gpu_model": "NVIDIA B200", "total_gpu_memory_gb": 1440, "maker": "NVIDIA", "headquarters": "Santa Clara, United States", "gpu_vendor": "NVIDIA", "cooling_type": null, "cpu": "2 × Intel Xeon Platinum 8570 (112 cores total)", "cpu_sockets": 2, "max_system_memory_tb": 4, "networking": "4 × OSFP serving 8 × single-port ConnectX-7 VPI; 2 × dual-port QSFP112 BlueField-3 DPU", "operating_temperature_c": "10–35" }, "sources": [ { "url": "https://www.nvidia.com/en-us/data-center/dgx-b200/", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "max_system_memory_tb", "networking", "operating_temperature_c", "system_power_kw" ] }, { "url": "https://docs.nvidia.com/dgx/dgxb200-user-guide/introduction-to-dgxb200.html", "accessed": "2026-11-11", "fields": [ "gpu_count", "gpu_model", "gpu_vendor", "power_supply", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://www.nvidia.com/en-us/contact/", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "nvidia-dgx-b300", "name": "NVIDIA DGX B300", "url": "https://www.nvidia.com/en-us/data-center/dgx-b300/", "summary": "NVIDIA's turnkey 10U system with eight Blackwell Ultra GPUs, 2.3 TB of GPU memory and eight 800 Gb/s ConnectX-8 ports. Built for enterprises and AI factories that want a ready-to-run training and inference node.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs" ], "fields": { "maker": "NVIDIA", "headquarters": "Santa Clara, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA B300 Blackwell Ultra", "gpu_count": 8, "rack_units": 10, "gpu_memory_gb": 288, "total_gpu_memory_gb": 2300, "cpu": "2 × Intel Xeon Platinum 6776P", "cpu_sockets": 2, "max_system_memory_tb": 4, "system_power_kw": 15, "power_supply": "12 × 3.2 kW; N+N redundancy", "networking": "8 × 800 Gb/s OSFP ConnectX-8 (InfiniBand/Ethernet); 2 × dual-port 400 Gb/s BlueField-3 DPUs", "operating_temperature_c": "10-35", "document_date": "2026-11-10" }, "sources": [ { "url": "https://docs.nvidia.com/dgx/dgxb300-user-guide/introduction-to-dgxb300.html", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "max_system_memory_tb", "networking", "operating_temperature_c", "power_supply", "rack_units", "system_power_kw", "total_gpu_memory_gb" ] }, { "url": "https://www.nvidia.com/en-us/contact/", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "nvidia-dgx-gb200-nvl72", "name": "NVIDIA DGX GB200 NVL72", "url": "https://www.nvidia.com/en-us/data-center/dgx-gb200/", "summary": "A complete NVLink-connected rack for large-scale AI training and inference. Liquid cold plates cool CPUs and GPUs while fans cool networking and storage components.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "cooling": "Liquid-cooled compute trays; air-cooled networking and storage", "rack_power_kw": 120, "gpu_count": 72, "gpu_model": "NVIDIA Blackwell", "form_factor": "Rack-scale; 18 compute trays of 1RU each", "maker": "NVIDIA", "headquarters": "Santa Clara, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://docs.nvidia.com/dgx/dgxgb200-user-guide/hardware.html", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "rack_power_kw" ] }, { "url": "https://www.nvidia.com/en-us/contact/", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.nvidia.com/content/dam/en-zz/Solutions/csr/NVIDIA_Corporation_CDP_Climate_Change_Questionnaire_2019.pdf", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "nvidia-dgx-gb300", "name": "NVIDIA DGX GB300", "url": "https://www.nvidia.com/en-us/data-center/dgx-gb300/", "summary": "A liquid-cooled, rack-scale DGX that links 72 Blackwell Ultra GPUs and 36 Grace CPUs in one NVLink domain. Built for enterprises and AI factories that want NVIDIA's own turnkey reasoning and training system.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "NVIDIA", "headquarters": "Santa Clara, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (GB300 Grace Blackwell Ultra Superchips)", "gpu_count": 72, "rack_cpu_count": 36, "cpu": "36 x NVIDIA Grace CPU (2,592 Arm Neoverse V2 cores)", "cooling_type": "Liquid", "cooling": "Rack-scale, liquid-cooled system", "form_factor": "Rack-scale, NVIDIA MGX rack", "interconnect": "NVIDIA NVLink Switch System with 9 x L1 NVLink Switches", "networking": "72 x OSFP single-port ConnectX-8 VPI (800 Gb/s InfiniBand); 18 x dual-port BlueField-3 VPI (200 Gb/s InfiniBand and Ethernet)" }, "sources": [ { "url": "https://www.nvidia.com/en-us/data-center/dgx-gb300/", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "networking", "rack_cpu_count" ] }, { "url": "https://www.nvidia.com/en-us/contact/", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "hpe-nvidia-gb300-nvl72", "name": "NVIDIA GB300 NVL72 by HPE", "url": "https://www.hpe.com/us/en/products/compute/hpe-nvidia-gb300-nvl72.html", "summary": "HPE's direct-liquid-cooled GB300 NVL72 rack with 72 Blackwell Ultra GPUs and 36 Grace CPUs, with in-row and in-rack CDU options. For hyperscale and sovereign AI builders who want HPE integration and services.", "hq": "US", "founded": 2015, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "HPE", "headquarters": "Spring, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (GB300 NVL72)", "gpu_count": 72, "rack_cpu_count": 36, "cpu": "36 x NVIDIA Grace CPU", "cooling_type": "Liquid", "cooling": "Direct liquid cooling; about 10% of heat captured by air and 90% by liquid; in-row and rack-mounted CDU options", "rack_power_kw": 140, "interconnect": "NVIDIA NVLink Switch, 72-GPU NVLink domain", "networking": "NVIDIA ConnectX-8 SuperNICs and NVIDIA InfiniBand" }, "sources": [ { "url": "https://www.hpe.com/us/en/products/compute/hpe-nvidia-gb300-nvl72.html", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "networking", "rack_cpu_count", "rack_power_kw" ] }, { "url": "https://www.hpe.com/int/en/about/enterprise-technology-us.html", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "pegatron-as205-2t1", "name": "PEGATRON AS205-2T1", "url": "https://svr.pegatroncorp.com/products/server/ai-gpu-server/as205-2t1", "summary": "A compact dual-socket NVIDIA MGX server for agentic AI, scientific computing, rendering, and video workloads. Its GPU and storage configurations vary with accelerator choice.", "hq": "TW", "founded": 2008, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air", "gpu_count": 4, "power_supply": "H200 NVL / RTX PRO 6000: 4 × 2400 W; 3+1 redundancy", "rack_units": 2, "gpu_model": "NVIDIA H200 NVL, RTX PRO 6000, or L40S", "maker": "Pegatron", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": "Air" }, "sources": [ { "url": "https://svr.pegatroncorp.com/specFile/AS205-2T1/PEGATRON_AS205-2T1_DS_2025v1.1.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_model", "gpu_vendor", "power_supply", "rack_units" ] }, { "url": "https://www.pegatroncorp.com/media/20250207032503169_2016CSR_EN.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.pegatroncorp.com/media/20250207032503169_2018CSR_EN.pdf", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "pegatron-as404-2a1-8i2", "name": "Pegatron AS404-2A1-8I2", "url": "https://www.storagenewsletter.com/2026/07/28/amd-aai-2026-pegatron-accelerates-enterprise-ai-at-scale-with-6th-gen-amd-epyc-server-cpus-and-amd-pensando-pollara-400-ai-nic", "summary": "A 4U liquid-cooled server with eight AMD Instinct MI355X GPUs and dual EPYC 9005 CPUs for AI training, inference and HPC.", "hq": "TW", "founded": 2008, "tags": [ "AMD GPUs", "Liquid-cooled" ], "fields": { "maker": "Pegatron", "gpu_vendor": "AMD", "gpu_model": "AMD Instinct MI355X", "gpu_count": 8, "rack_units": 4, "cooling_type": "Liquid", "cooling": "Advanced liquid cooling", "cpu": "Dual AMD EPYC 9005 Series", "cpu_sockets": 2, "networking": "AMD Pensando Pollara 400 AI NIC supported", "headquarters": "Taipei, Taiwan" }, "sources": [ { "url": "https://www.storagenewsletter.com/2026/07/28/amd-aai-2026-pegatron-accelerates-enterprise-ai-at-scale-with-6th-gen-amd-epyc-server-cpus-and-amd-pensando-pollara-400-ai-nic", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_model", "gpu_vendor", "networking", "rack_units" ] }, { "url": "https://www.pegatroncorp.com/media/20250207032503169_2016CSR_EN.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "penguin-altus-xe4318gts-dtc", "name": "Penguin Solutions Altus XE4318GTS-DTC", "url": "https://www.penguinsolutions.com/en-us/products/altus-amd-servers", "summary": "An AMD EPYC HGX B200 server for AI training, inference, and HPC. Direct liquid cooling supports a compact chassis.", "hq": "US", "founded": 1988, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "cooling": "Liquid", "gpu_memory_gb": 180, "gpu_count": 8, "rack_units": 4, "gpu_model": "NVIDIA HGX B200", "cpu": "AMD EPYC 9005/9004", "maker": "Penguin Solutions", "headquarters": "Fremont, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://www.penguinsolutions.com/en-us/products/gpu-accelerated-servers", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "rack_units" ] }, { "url": "https://www.penguinsolutions.com/en-us/company/ess/environment", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.penguinsolutions.com/en-us/company/careers", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "qct-quantagrid-d75h-10u", "name": "QCT QuantaGrid D75H-10U", "url": "https://www.qct.io/product/index/Server/rackmount-server/GPGPU-Xeon-Phi/QuantaGrid-D75H-10U", "summary": "QCT's 10U HGX B300 server with two Xeon 6 processors, 12 front-access PSUs and eight 800G ConnectX-8 ports. Built for large AI training and inference deployments.", "hq": "TW", "founded": 2012, "tags": [ "NVIDIA GPUs" ], "fields": { "maker": "QCT (Quanta)", "headquarters": "Taoyuan, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 10, "cpu": "2 × Intel Xeon 6", "cpu_sockets": 2, "max_system_memory_tb": 8, "memory_slots": 32, "power_supply": "12 × 3200 W; N+N redundancy", "networking": "8 × ConnectX-8 OSFP, 800G Ethernet or InfiniBand; BlueField-3 DPU support", "operating_temperature_c": "5-35", "document_date": "2025-10-23" }, "sources": [ { "url": "https://www.qct.io/product/index/Server/rackmount-server/GPGPU-Xeon-Phi/QuantaGrid-D75H-10U", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://go.qct.io/wp-content/uploads/2022/04/QCT-Telco-Product-Portfolio_2022FEB.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "qct-quantagrid-d75l-3u", "name": "QCT QuantaGrid D75L-3U", "url": "https://www.qct.io/product/index/Server/rackmount-server/GPGPU-Xeon-Phi/QuantaGrid-D75L-3U", "summary": "QCT's 3U liquid-cooled HGX B300 node for 19-inch MGX racks with a busbar and power shelves, up to nine nodes (72 GPUs) per rack.", "hq": "TW", "founded": 2012, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "QCT (Quanta)", "headquarters": "Taoyuan, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 3, "cooling_type": "Liquid", "cooling": "Advanced liquid-cooling technology; MGX cooling manifolds", "cpu": "2 × Intel Xeon 6700-series (P-core)", "cpu_sockets": 2, "total_gpu_memory_gb": 2300, "memory_slots": 32, "networking": "8 × ConnectX-8 SuperNICs, up to 800Gb/s", "document_date": "2026-07-22" }, "sources": [ { "url": "https://www.qct.io/product/index/Server/rackmount-server/GPGPU-Xeon-Phi/QuantaGrid-D75L-3U", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "memory_slots", "networking", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://go.qct.io/wp-content/uploads/2022/04/QCT-Telco-Product-Portfolio_2022FEB.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "qct-quantagrid-d75t-7u", "name": "QCT QuantaGrid D75T-7U", "url": "https://www.qct.io/", "summary": "An AMD EPYC-based GPU platform from Quanta’s data-center business. AMD’s partner catalog identifies air-cooled MI325X and MI350X configurations.", "hq": "TW", "founded": 2012, "tags": [ "AMD GPUs", "Air-cooled" ], "fields": { "networking": "AMD Pensando Pollara-1Q 400P (HHHL)", "cooling": "Air", "rack_units": 7, "gpu_model": "AMD Instinct MI325X or MI350X", "cpu": "AMD EPYC 9005", "maker": "QCT (Quanta)", "headquarters": "Taoyuan, Taiwan", "gpu_vendor": "AMD", "cooling_type": "Air" }, "sources": [ { "url": "https://www.amd.com/content/dam/amd/en/documents/pensando-technical-docs/other/pensando-pollara-400-ai-nic-ready-server-platforms-catalog.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_model", "gpu_vendor", "networking", "rack_units" ] }, { "url": "https://go.qct.io/wp-content/uploads/2022/04/QCT-Telco-Product-Portfolio_2022FEB.pdf", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.prweb.com/releases/quanta_qct_simplifies_clarifies_brand_with_name_change_to_quanta_cloud_technology_qct_/prweb12047986.htm", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-as-4126gs-nbr-lcc", "name": "Supermicro A+ Server AS-4126GS-NBR-LCC", "url": "https://www.supermicro.com/en/products/system/gpu/4u/as-4126gs-nbr-lcc", "summary": "A liquid-cooled AMD EPYC HGX B200 server for AI and scientific computing. Supermicro requires a complete rack solution and onsite service with this system.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "cooling": "Direct-to-chip cold plate liquid cooling", "gpu_memory_gb": 180, "gpu_count": 8, "rack_units": 4, "gpu_model": "NVIDIA HGX B200 SXM", "max_system_memory_tb": 6, "memory_slots": 24, "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Liquid" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/4u/as-4126gs-nbr-lcc", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-as-4126gs-nmr-lcc", "name": "Supermicro AS-4126GS-NMR-LCC", "url": "https://www.supermicro.com/en/products/system/gpu/4u/as-4126gs-nmr-lcc", "summary": "A direct-liquid-cooled 4U dual-EPYC server with eight OAM accelerators for dense AI racks. Takes AMD Instinct MI355X or MI325X.", "hq": "US", "founded": 1993, "tags": [ "AMD GPUs", "Liquid-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "AMD", "gpu_model": "AMD Instinct MI355X (one supported option)", "gpu_count": 8, "rack_units": 4, "form_factor": "4U rackmount", "cooling_type": "Liquid", "cooling": "Direct-to-chip liquid cold plates, up to 5 × 6 cm fans and 1 air shroud", "cpu": "Dual AMD EPYC 9005/9004 Series, up to 384 cores", "cpu_sockets": 2, "max_system_memory_tb": 6, "memory_slots": 24, "power_supply": "4 × 6600 W Titanium (96%); 2+2 redundancy", "interconnect": "AMD Infinity Fabric Link (GPU-to-GPU); PCIe 5.0 x16 (CPU-to-GPU)", "operating_temperature_c": "10 to 35" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/4u/as-4126gs-nmr-lcc", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "max_system_memory_tb", "memory_slots", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-as-8126gs-tnmr", "name": "Supermicro AS-8126GS-TNMR", "url": "https://www.supermicro.com/en/products/system/gpu/10u/as-8126gs-tnmr", "summary": "An air-cooled 8U dual-EPYC server with eight OAM accelerators, built for AI training and inference in standard air-cooled racks. Takes AMD Instinct MI350X or MI325X.", "hq": "US", "founded": 1993, "tags": [ "AMD GPUs", "Air-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "AMD", "gpu_model": "AMD Instinct MI350X (one supported option)", "gpu_count": 8, "rack_units": 8, "form_factor": "8U rackmount", "cooling_type": "Air", "cooling": "Air-cooled, up to 14 heavy-duty fans and 1 air shroud", "cpu": "Dual AMD EPYC 9005/9004 Series, up to 384 cores", "cpu_sockets": 2, "max_system_memory_tb": 6, "memory_slots": 24, "power_supply": "6 × 5250 W Titanium (96%); 3+3 redundancy; 200-240 Vac", "interconnect": "AMD Infinity Fabric Link (GPU-to-GPU); PCIe 5.0 x16 (CPU-to-GPU)", "operating_temperature_c": "10 to 35" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/10u/as-8126gs-tnmr", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "max_system_memory_tb", "memory_slots", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-822gs-nbrt", "name": "Supermicro GPU SuperServer SYS-822GS-NBRT", "url": "https://www.supermicro.com/en/products/system/gpu/8u/sys-822gs-nbrt", "summary": "An eight-GPU B200 server for enterprise AI training, inference, and scientific computing. The front-access chassis uses air cooling.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "cooling": "Air cooled", "gpu_count": 8, "power_supply": "6 × 6600 W; 3+3 redundancy", "rack_units": 8, "gpu_model": "NVIDIA HGX B200 SXM6", "max_system_memory_tb": 8, "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "cooling_type": "Air", "gpu_memory_gb": 180, "gpu_tdp_w": 1000, "memory_slots": 32, "cpu": "Dual Intel Xeon 6700/6500 series with P-cores, up to 350 W TDP", "cpu_sockets": 2 }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/datasheet/sys-822gs-nbrt", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor", "max_system_memory_tb", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-srs-gb200-nvl72", "name": "Supermicro SRS-GB200-NVL72", "url": "https://www.supermicro.com/en/products/rack/srs-gb200-nvl72", "summary": "Liquid-cooled rack with 72 Blackwell GPUs and 36 Grace CPUs on one NVLink domain, delivered as an integrated Supermicro rack. For AI factories that want exascale compute per rack.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA GB200 NVL72", "gpu_count": 72, "rack_cpu_count": 36, "cooling_type": "Liquid", "cooling": "Fully liquid-cooled: 1 × in-rack 4U 250 kW CDU (redundant PSU, dual hot-swap pumps), 1.3 MW in-row CDU, or optional 180/240 kW liquid-to-air units", "cpu": "36 × NVIDIA Grace, 72-core Arm Neoverse V2", "rack_power_kw": 132, "rack_power_delivery": "8 × 1U 33 kW power shelves (6 × 5.5 kW PSUs each), 132 kW total; operating power 125 to 135 kW", "interconnect": "9 × NVLink switch trays; 1.8 TB/s GPU-to-GPU", "document_date": "2025-02" }, "sources": [ { "url": "https://www.supermicro.com/datasheet/datasheet_SuperCluster_GB200_NVL72.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "document_date", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "rack_cpu_count", "rack_power_delivery", "rack_power_kw" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-srs-gb300-nvl72", "name": "Supermicro SuperServer SRS-GB300-NVL72", "url": "https://www.supermicro.com/en/products/system/gpu/48u/srs-gb300-nvl72", "summary": "Supermicro's liquid-cooled 48U GB300 NVL72 rack: 18 x 1U compute trays, 72 B300 GPUs, and an in-rack 250 kW CDU. For AI factories that want a rack delivered integrated and tested.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA B300 (GB300 NVL72)", "gpu_count": 72, "gpu_memory_gb": 288, "rack_cpu_count": 36, "cpu": "36 x NVIDIA Grace CPU", "rack_units": 48, "form_factor": "Rack-scale, 48U 19-inch rack with 18 x 1U ARS-121GL-NB3 trays", "max_system_memory_tb": 17, "cooling_type": "Liquid", "cooling": "Direct liquid cooling with one in-rack 250 kW CDU; in-row CDU (up to 1.8 MW) and L2A sidecar CDU (up to 200 kW) options", "rack_power_kw": 132, "rack_power_delivery": "8 x 1U power shelves, each 33 kW (6 x 5.5 kW PSUs)", "interconnect": "9 x NVLink Switch, 1.8 TB/s GPU-to-GPU bandwidth", "networking": "Up to 800 Gb/s NVIDIA Quantum-X800 InfiniBand or Spectrum-X Ethernet with ConnectX-8 SuperNICs; BlueField-3 DPUs up to 200 Gb/s in-band management" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/48u/srs-gb300-nvl72", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "form_factor", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "interconnect", "max_system_memory_tb", "networking", "rack_cpu_count", "rack_power_delivery", "rack_power_kw", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-222gs-nb3ot-alc", "name": "Supermicro SYS-222GS-NB3OT-ALC", "url": "https://www.supermicro.com/manuals/brochure/brochure_NVIDIA-Blackwell-Solutions.pdf", "summary": "Supermicro's 2-OU liquid-cooled HGX B300 node for 21-inch OCP Open Rack V3, with nine nodes per 72-GPU rack. Built for hyperscale-style B300 deployments.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "form_factor": "2 OU, 21-inch OCP ORV3", "cooling_type": "Liquid", "cpu": "2 × Intel Xeon 6700 series (P-core)", "cpu_sockets": 2, "gpu_memory_gb": 288, "gpu_tdp_w": 1100, "memory_slots": 32, "max_system_memory_tb": 8, "networking": "8 integrated NVIDIA ConnectX-8 SuperNICs, up to 800Gb/s", "rack_power_delivery": "8 × 1U 33 kW power shelves (6 × 5.5 kW PSUs each), total 132 kW, N+N redundancy (72-GPU, 44-OU rack)" }, "sources": [ { "url": "https://www.supermicro.com/manuals/brochure/brochure_NVIDIA-Blackwell-Solutions.pdf", "accessed": "2026-11-11", "fields": [ "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_tdp_w", "gpu_vendor", "max_system_memory_tb", "memory_slots", "networking", "rack_power_delivery" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-422gs-nb3rt-alc", "name": "Supermicro SYS-422GS-NB3RT-ALC", "url": "https://www.supermicro.com/en/products/system/gpu/4u/sys-422gs-nb3rt-alc", "summary": "Supermicro's 4U direct-to-chip liquid-cooled HGX B300 server, sold with full liquid-cooled rack integration. Built for dense B300 racks.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 4, "cooling_type": "Liquid", "cooling": "Direct-to-chip liquid cooling (cold plates), up to 95% heat capture at max configuration", "cpu": "2 × Intel Xeon 6700 series (P-core)", "cpu_sockets": 2, "gpu_memory_gb": 288, "total_gpu_memory_gb": 2300, "max_system_memory_tb": 8, "memory_slots": 32, "power_supply": "4 × 6600 W; 2+2 redundancy, Titanium", "networking": "8 × 800Gb/s OSFP ConnectX-8 SuperNIC; 2 × 10GbE RJ45" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/4u/sys-422gs-nb3rt-alc", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "networking", "power_supply", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-422gs-nbrt-lcc", "name": "Supermicro SYS-422GS-NBRT-LCC", "url": "https://www.supermicro.com/en/products/system/gpu/4u/sys-422gs-nbrt-lcc", "summary": "4U direct-liquid-cooled server with eight NVIDIA HGX B200 GPUs and dual Xeon 6700 CPUs. Sold with full liquid-cooled rack integration for dense AI deployments.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Liquid-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B200", "gpu_count": 8, "rack_units": 4, "form_factor": "4U rackmount", "cooling_type": "Liquid", "cooling": "Direct-to-chip liquid cold plates, up to 87% heat capture in maximum configuration; 4 × 80 mm fans for remaining air cooling", "cpu": "Dual Intel Xeon 6700-series with P-cores, up to 86 cores each", "cpu_sockets": 2, "gpu_memory_gb": 180, "total_gpu_memory_gb": 1440, "max_system_memory_tb": 4, "memory_slots": 32, "power_supply": "4 × 6600 W Titanium (96%); 2+2 redundancy", "interconnect": "NVLink with NVSwitch", "networking": "2 × 10GbE RJ45 (Intel X710-AT2), 1 × dedicated BMC/IPMI", "operating_temperature_c": "10 to 35" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/4u/sys-422gs-nbrt-lcc", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "interconnect", "max_system_memory_tb", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-822ga-ngr3", "name": "Supermicro SYS-822GA-NGR3", "url": "https://www.supermicro.com/en/products/system/ai/8u/sys-822ga-ngr3", "summary": "8U air-cooled AI training server with eight Intel Gaudi 3 OAM accelerators and dual Xeon 6900 CPUs. Built for teams that want Gaudi 3 scale-out over six onboard 800 GbE ports.", "hq": "US", "founded": 1993, "tags": [ "Intel Gaudi", "Air-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "Intel", "gpu_model": "Intel Gaudi 3 (OAM)", "gpu_count": 8, "rack_units": 8, "form_factor": "8U rackmount", "cooling_type": "Air", "cooling": "Air-cooled, up to 10 removable 80 x 80 x 80 mm fans", "cpu": "Dual Intel Xeon 6900-series with P-cores, up to 128 cores each", "cpu_sockets": 2, "max_system_memory_tb": 6, "memory_slots": 24, "power_supply": "8 × 3000 W Titanium (96%); 4+4 redundancy", "interconnect": "24 × 200 GbE (48 × 112 Gbps PAM4 SerDes) GPU-to-GPU", "networking": "6 × OSFP 800 GbE onboard", "operating_temperature_c": "10 to 35" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/ai/8u/sys-822ga-ngr3", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "max_system_memory_tb", "memory_slots", "networking", "operating_temperature_c", "power_supply", "rack_units" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "supermicro-sys-822gs-nb3rt", "name": "Supermicro SYS-822GS-NB3RT", "url": "https://www.supermicro.com/en/products/system/gpu/8u/sys-822gs-nb3rt", "summary": "Supermicro's 8U air-cooled HGX B300 server with dual Xeon 6700 and eight 800GbE ConnectX-8 ports. Built for data centers that deploy B300 without liquid cooling.", "hq": "US", "founded": 1993, "tags": [ "NVIDIA GPUs", "Air-cooled" ], "fields": { "maker": "Supermicro", "headquarters": "San Jose, United States", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA HGX B300", "gpu_count": 8, "rack_units": 8, "cooling_type": "Air", "cooling": "Air cooled, up to 12 heavy-duty fans", "cpu": "2 × Intel Xeon 6700 series (P-core)", "cpu_sockets": 2, "gpu_memory_gb": 288, "total_gpu_memory_gb": 2300, "max_system_memory_tb": 8, "memory_slots": 32, "system_power_kw": 15, "power_supply": "6 × 6600 W; 3+3 redundancy, Titanium", "networking": "8 × 800GbE OSFP ConnectX-8 SuperNIC; 2 × 10GbE RJ45" }, "sources": [ { "url": "https://www.supermicro.com/en/products/system/gpu/8u/sys-822gs-nb3rt", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "cpu_sockets", "gpu_count", "gpu_memory_gb", "gpu_model", "gpu_vendor", "max_system_memory_tb", "memory_slots", "networking", "power_supply", "rack_units", "total_gpu_memory_gb" ] }, { "url": "https://www.supermicro.com/datasheet/Datasheet_Supermicro_NVIDIA_AI_Factories_HGX_B300.pdf", "accessed": "2026-11-11", "fields": [ "system_power_kw" ] }, { "url": "https://www.supermicro.com/en/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "thinkmate-gpx-ge95-2808-8hgx-b2", "name": "Thinkmate GPX GE95-2808-8HGX-B2", "url": "https://www.thinkmate.com/system/gpx-ge95-2808-8hgx-b2", "summary": "A configurable AMD EPYC HGX B200 system for AI and accelerated-computing buyers. Thinkmate assembles and tests configurations and offers advanced parts-replacement warranty options.", "hq": "US", "founded": 1986, "tags": [ "NVIDIA GPUs" ], "fields": { "gpu_count": 8, "warranty": "Five-year advanced parts replacement option", "power_supply": "12 × 3000 W; 6+6 redundancy", "rack_units": 8, "gpu_model": "NVIDIA HGX B200 SXM", "total_gpu_memory_gb": 1440, "maker": "Thinkmate", "headquarters": "Norwood, United States", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://www.thinkmate.com/system/gpx-ge95-2808-8hgx-b2", "accessed": "2026-11-11", "fields": [ "gpu_count", "gpu_model", "gpu_vendor", "power_supply", "rack_units", "total_gpu_memory_gb", "warranty" ] }, { "url": "https://www.thinkmate.com/about", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "wistron-gs-18g00", "name": "Wistron GS-18G00", "url": "https://www.wistron.com/", "summary": "An AMD-based AI server platform from the Taiwanese ODM. AMD lists it among partner platforms supporting the Pollara 400 AI network adapter.", "hq": "TW", "founded": 2001, "tags": [ "AMD GPUs", "Air-cooled" ], "fields": { "networking": "AMD Pensando Pollara-1Q 400P (HHHL)", "cooling": "Air", "rack_units": 8, "gpu_model": "AMD Instinct MI350X", "cpu": "AMD EPYC 9005", "maker": "Wistron", "headquarters": "Taiwan", "gpu_vendor": "AMD", "cooling_type": "Air" }, "sources": [ { "url": "https://www.amd.com/content/dam/amd/en/documents/pensando-technical-docs/other/pensando-pollara-400-ai-nic-ready-server-platforms-catalog.pdf", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "cpu", "gpu_model", "gpu_vendor", "networking", "rack_units" ] }, { "url": "https://jobs.wistron.com/career/en-us/about-us?site=hq", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "wiwynn-nvidia-gb300-nvl72", "name": "Wiwynn NVIDIA GB300 NVL72", "url": "https://www.wiwynn.com/news/wiwynn-showcases-ai-servers-featuring-nvidia-gb300-nvl72-platform-and-liquid-cooling-innovations-at-gtc-2025", "summary": "Wiwynn's fully liquid-cooled, fan-less GB300 NVL72 AI rack for hyperscale data centers, with in-rack CDU and sidecar options.", "hq": "TW", "founded": 2012, "tags": [ "NVIDIA GPUs", "Liquid-cooled", "Rack-scale" ], "fields": { "maker": "Wiwynn", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "gpu_model": "NVIDIA Blackwell Ultra (GB300 NVL72)", "cooling_type": "Liquid", "cooling": "Fully liquid cooling design, fan-less rack", "document_date": "2025-03-18" }, "sources": [ { "url": "https://www.wiwynn.com/news/wiwynn-showcases-ai-servers-featuring-nvidia-gb300-nvl72-platform-and-liquid-cooling-innovations-at-gtc-2025", "accessed": "2026-11-11", "fields": [ "cooling", "cooling_type", "document_date", "gpu_model", "gpu_vendor" ] }, { "url": "https://www.wiwynn.com/about-wiwynn/global-operations", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" }, { "id": "wiwynn-nvidia-vera-rubin-nvl72", "name": "Wiwynn NVIDIA Vera Rubin NVL72", "url": "https://www.wiwynn.com/solutions/ai-infrastructure", "summary": "Wiwynn integrates rack-scale NVIDIA systems for hyperscale cloud providers. Its Vera Rubin offering uses the third-generation MGX framework.", "hq": "TW", "founded": 2012, "tags": [ "NVIDIA GPUs", "Rack-scale" ], "fields": { "interconnect": "Cable-less interconnect topology", "gpu_count": 72, "gpu_model": "NVIDIA Rubin", "form_factor": "Rack-scale", "rack_cpu_count": 36, "maker": "Wiwynn", "headquarters": "Taipei, Taiwan", "gpu_vendor": "NVIDIA", "cooling_type": null }, "sources": [ { "url": "https://www.wiwynn.com/solutions/ai-infrastructure", "accessed": "2026-11-11", "fields": [ "form_factor", "gpu_count", "gpu_model", "gpu_vendor", "interconnect", "rack_cpu_count" ] }, { "url": "https://www.wiwynn.com/about-wiwynn/global-operations", "accessed": "2026-11-11", "fields": [ "headquarters" ] }, { "url": "https://www.wiwynn.com/about-wiwynn/company-info", "accessed": "2026-11-11", "fields": [] } ], "verifiedAt": "2026-11-11" }, { "id": "xfusion-fusionserver-g8600-v7", "name": "xFusion FusionServer G8600 V7", "url": "https://www.xfusion.com/en/product/heterogeneous-server/fusionserver-g8600-v7", "summary": "An eight-GPU platform for model training, HPC, and data analysis. Separate GPU and CPU power planes support modular servicing.", "hq": "CN", "founded": 2021, "tags": [], "fields": { "cpu_sockets": 2, "gpu_count": 8, "power_supply": "54 V PSUs with N+M redundancy", "rack_units": 8, "memory_slots": 32, "cpu": "4th/5th Gen Intel Xeon Scalable", "maker": "xFusion", "headquarters": "Zhengzhou, China", "gpu_vendor": null, "cooling_type": null }, "sources": [ { "url": "https://www.xfusion.com/en/product/heterogeneous-server/fusionserver-g8600-v7", "accessed": "2026-11-11", "fields": [ "cpu", "cpu_sockets", "gpu_count", "memory_slots", "power_supply", "rack_units" ] }, { "url": "https://www.xfusion.com/en/news/xfusion-explorer-2026", "accessed": "2026-11-11", "fields": [] }, { "url": "https://www.xfusion.com/cn/news/xfusion-strategic-press-conference2025", "accessed": "2026-11-11", "fields": [ "headquarters" ] } ], "verifiedAt": "2026-11-11" } ] }